Search Azure terms

AI and Machine Learning premium Grounded generation Grounded generation is a generative AI pattern where model responses are based on supplied source material, retrieved enterprise content, tool results, or other approved grounding data. Generative AI intermediate 4 commands Aliases: grounded AI generation, grounded response generation, RAG answer generation Quick peek Open full term page

AI and Machine Learning premium Groundedness detection Groundedness detection checks whether a generated text response is supported by provided source material and flags responses that appear ungrounded or inconsistent with that material. Azure AI Content Safety advanced 4 commands Aliases: groundedness filter, ungroundedness detection, AI groundedness check Quick peek Open full term page

AI and Machine Learning template-specs-upgraded Retrieval augmented generation Microsoft Learn describes retrieval augmented generation as an AI pattern where a user query retrieves relevant grounding content, often from Azure AI Search or Foundry indexes, before a model generates an answer. The goal is more accurate, cited, and current responses from private or changing data. Grounded generation fundamentals 5 commands Aliases: RAG, retrieval augmented generation, grounded generation, Azure AI Search RAG, agentic RAG Quick peek Open full term page

AI and Machine Learning premium Input token Input token controls how much information an AI request can send to the model and how that request consumes quota, cost, latency, and context capacity. Teams see it in azure openai requests, chat completions. It is not an output token, API key, authentication token, session token, or tokenizer algorithm; confusing them can create truncated prompts, high cost. Use the term when reviewing access, monitoring, cost, recovery, or performance. It keeps architects, operators, security reviewers, and support teams focused on the same setting, resource, or behavior. Azure OpenAI Fundamentals 5 commands Aliases: prompt token, input tokens, request tokens, model input token, context input token Quick peek Open full term page

AI and Machine Learning premium Grounding Grounding is the practice of supplying relevant external information, retrieved content, or tool results so a generative AI model can base its answer on known context. AI platform and search intermediate 4 commands Aliases: AI grounding, model grounding, grounding a prompt Quick peek Open full term page

AI and Machine Learning premium Grounding data Grounding data is the source content, retrieved documents, indexed chunks, tool results, or external information supplied to a model so generated output can be based on approved facts. Azure OpenAI intermediate 4 commands Aliases: RAG grounding data, source grounding data, AI source material Quick peek Open full term page

AI and Machine Learning verified RAG Retrieval-augmented generation grounds generative AI responses in retrieved enterprise content instead of relying only on model training. In Azure, RAG commonly combines Azure AI Search indexes, grounding data, embeddings, and an Azure OpenAI or Foundry model to answer using current, domain-specific information. Generative AI advanced 5 commands Aliases: Retrieval augmented generation, retrieval-augmented generation Quick peek Open full term page

AI and Machine Learning premium Image generation model Image generation model is an Azure OpenAI or Foundry model deployment that creates images from prompts, commonly using the current gpt-image model family. Azure OpenAI intermediate 4 commands Aliases: Azure OpenAI image model, gpt-image model, image generation model, Image generation model, text-to-image model Quick peek Open full term page

AI and Machine Learning premium Jailbreak detection Jailbreak detection is the AI safety capability that identifies prompts or embedded document instructions attempting to bypass system rules, policies, or intended model behavior. AI safety intermediate 5 commands Aliases: Prompt Shields, prompt attack detection, jailbreak risk detection, user prompt attack detection, document attack detection Quick peek Open full term page

AI and Machine Learning verified RAG evaluation RAG evaluation measures whether a retrieval-augmented system finds relevant grounding documents and produces answers that are grounded, relevant, coherent, and safe. Azure AI Foundry includes RAG evaluators for assessing retrieval quality and generation quality before releasing or changing an AI experience. Microsoft Foundry intermediate 5 commands Aliases: retrieval augmented generation evaluation, RAG evaluators Quick peek Open full term page

AI and Machine Learning verified RAG pipeline A RAG pipeline is the end-to-end flow that ingests content, chunks and enriches it, indexes it for retrieval, sends retrieved context to a model, and evaluates responses. In Azure, it often connects Azure AI Search, Azure OpenAI, storage, identity, monitoring, and evaluation tooling. Azure OpenAI advanced 5 commands Aliases: retrieval augmented generation pipeline, RAG workflow Quick peek Open full term page

AI and Machine Learning premium Azure OpenAI Service Azure OpenAI Service provides managed access to OpenAI model capabilities through Azure endpoints and enterprise controls. Generative AI fundamentals 11 commands Aliases: Azure OpenAI Quick peek Open full term page

AI and Machine Learning premium Azure OpenAI Azure OpenAI provides OpenAI model capabilities through Azure resources, deployments, identity, networking, quota, and monitoring controls. Azure OpenAI intermediate 8 commands Aliases: Azure OpenAI Service Quick peek Open full term page

Databases premium Cosmos DB vector embedding policy Cosmos DB vector embedding policy is the container policy that defines vector paths, dimensions, data type, and distance function for embeddings stored in Cosmos DB for NoSQL. It tells Cosmos DB which item properties contain embeddings and how those vectors should be interpreted. You see it when teams build semantic search, recommendation, image similarity, or retrieval-augmented generation features on Cosmos DB data. The production check is whether the vector path, dimensions, data type, and distance function are correct before production data lands. Document the decision in code, templates, metrics, and runbooks. Azure Cosmos DB intermediate 7 commands Aliases: No aliases yet Quick peek Open full term page

Containers premium Event-driven job An Event-driven job in Azure Container Apps starts job executions when a configured scale rule detects work from an event source. Teams use it to run finite container work such as queue processing, batch enrichment, report generation, or cleanup only when events or messages are waiting. It is not a continuously running container app, an AKS deployment, a scheduled job, or a guarantee that every event is processed exactly once. In production, confirm job trigger type, scale rule, minimum and maximum executions, polling interval, image version, secret references, managed identity, execution status, logs, retries, and queue depth before treating the. Azure Container Apps intermediate 6 commands Aliases: Container Apps event-driven job, KEDA job, event-triggered container job Quick peek Open full term page

Web premium Activity function Activity function is the worker function called by a Durable Functions orchestrator to perform one concrete unit of work. In everyday Azure work, teams use it to move side effects such as API calls, database writes, document generation, or messages out of deterministic orchestration code. The useful evidence is activity name, orchestration instance ID, retry Durable Functions intermediate 5 commands Aliases: Durable Functions activity, Durable activity function, orchestrator activity Quick peek Open full term page

Web premium App Service Environment v3 App Service Environment v3 is the current App Service Environment generation for single-tenant, dedicated App Service hosting with simplified networking and high-scale isolated workloads. App Service isolation intermediate 5 commands Aliases: Azure App Service Environment v3, app service environment v3 Quick peek Open full term page

Compute premium Gallery image definition Gallery image definition is the Azure Compute Gallery record that defines an image family before individual image versions are published. Teams use it to standardize OS type, generation, publisher, offer, SKU, architecture, and image state so teams deploy consistent virtual machine builds. In daily Azure work, it appears when engineers create hardened Windows or Linux baselines, separate application images, control VM generation compatibility, or publish reusable images across regions. It is not the actual disk snapshot, a VM instance, an image version, a patching process, or proof that every VM deployed from it is secure. Disks and images intermediate 5 commands Aliases: Azure Compute Gallery image definition, SIG image definition, image definition Quick peek Open full term page

Databases premium Instance pool Instance pool controls how multiple small Azure SQL Managed Instances share capacity for migrations that need instance-level compatibility with lower per-instance overhead. Teams see it in azure sql managed instance pools, virtual clusters. It is not an Azure SQL elastic pool, a VM scale set, a Hyperscale database, or a shared SQL Server instance; confusing them can create capacity exhaustion, subnet design mistakes. Use the term when reviewing access, monitoring, cost, recovery, or performance. It keeps architects, operators, security reviewers, and support teams focused on the same setting, resource, or behavior. Azure SQL Managed Instance Intermediate 5 commands Aliases: SQL Managed Instance pool, Azure SQL instance pool, managed instance pool, SQL MI instance pool Quick peek Open full term page

AI and Machine Learning premium Integrated vectorization Integrated vectorization controls how Azure AI Search creates and uses embeddings inside indexing and query pipelines so applications can run vector or hybrid search without external embedding glue code. Teams see it in azure ai search indexes, skillsets. It is not manual offline embedding generation, plain keyword search, semantic ranking alone, a vector index without a vectorizer, or a model fine-tuning job; confusing them can create dimension mismatches, failed indexer runs. Use the term when reviewing access, monitoring, cost, recovery, or performance. It keeps architects, operators, security reviewers, and support teams focused on the same setting, resource, or behavior. Azure AI Search Advanced 5 commands Aliases: Azure AI Search integrated vectorization, query-time vectorization, indexer vectorization, vectorizer, embedding skill Quick peek Open full term page

Web premium Kudu console Kudu console is the App Service Advanced Tools command console used to inspect runtime files, environment state, processes, logs, and deployment artifacts for a web app or function app. App Service diagnostics Intermediate 5 commands Aliases: Kudu debug console, App Service advanced tools console, SCM console, Kudu Advanced Tools Quick peek Open full term page

Analytics premium Kusto cache policy Kusto cache policy controls which Azure Data Explorer data is kept in hot local cache versus colder storage for faster queries and managed cost. Azure Data Explorer policies Advanced 5 commands Aliases: ADX cache policy, hot cache policy, Kusto hot cache, Azure Data Explorer caching policy Quick peek Open full term page

Analytics premium Kusto cluster Kusto cluster is the Azure Data Explorer compute resource that hosts databases, ingestion, query processing, scale settings, networking, and monitoring for Kusto workloads. Azure Data Explorer infrastructure Intermediate 5 commands Aliases: Azure Data Explorer cluster, ADX cluster, Microsoft.Kusto cluster Quick peek Open full term page

Analytics premium Kusto continuous export Kusto continuous export periodically runs a Kusto query and writes results to an external table destination such as Azure Storage for downstream processing or archiving. Azure Data Explorer export Advanced 5 commands Aliases: ADX continuous export, Kusto export policy, continuous data export Quick peek Open full term page

Analytics premium Kusto data connection Kusto data connection links Azure Data Explorer to event or storage sources such as Event Hubs, Event Grid, or IoT Hub so data can be ingested into a database table. Azure Data Explorer ingestion Intermediate 5 commands Aliases: ADX data connection, Event Hubs data connection, Event Grid data connection, IoT Hub data connection Quick peek Open full term page

Analytics premium Kusto database Kusto database is a logical container inside an Azure Data Explorer cluster that holds tables, functions, policies, permissions, and ingestion/query configuration. Azure Data Explorer database Intermediate 5 commands Aliases: ADX database, Azure Data Explorer database, Kusto DB Quick peek Open full term page

Analytics premium Kusto follower database Kusto follower database is a read-only database attached from another Azure Data Explorer cluster so consumers can query leader data without copying it. Azure Data Explorer data sharing Advanced 5 commands Aliases: ADX follower database, Azure Data Explorer follower database, followed database Quick peek Open full term page

Analytics premium Kusto function Kusto function is a reusable KQL query or query fragment, stored as a database entity or defined ad hoc, that standardizes analytics logic. Kusto query objects Intermediate 5 commands Aliases: Kusto stored function, ADX stored function, KQL function Quick peek Open full term page

Analytics premium Kusto ingestion Kusto ingestion is the process of loading data into Azure Data Explorer tables, where data is validated, mapped, indexed, encoded, and made queryable. Azure Data Explorer ingestion Intermediate 5 commands Aliases: ADX ingestion, Azure Data Explorer ingestion, Kusto data ingestion Quick peek Open full term page

Analytics premium Kusto materialized view Kusto materialized view is a managed aggregation over a source table or another materialized view that keeps a continuously updated result for faster queries. Kusto query optimization Advanced 5 commands Aliases: ADX materialized view, Kusto MV, materialized view in Azure Data Explorer Quick peek Open full term page

AI and Machine Learning premium Search index projection A search index projection defines how enriched content from an Azure AI Search skillset is projected into one or more indexes. It is commonly used to create chunked child documents with parent metadata for retrieval-augmented generation. It is configured as part of the enrichment pipeline. Azure AI Search fundamentals 5 commands Aliases: Azure AI Search index projection, indexer projection, projected search index, RAG chunk projection, skillset index projection, index projections, projection selector, projected index, chunk projection Quick peek Open full term page

AI and Machine Learning premium Search skill A search skill is one processing step inside an Azure AI Search skillset. It transforms extracted source content during indexing, such as OCR, entity recognition, translation, text splitting, embedding generation, or custom enrichment, and writes output into the enriched document tree. Search intermediate 5 commands Aliases: Azure AI Search skill, enrichment skill, cognitive skill, custom skill, indexing skill Quick peek Open full term page

AI and Machine Learning premium Agent run Agent run is one execution of an agent against a thread, during which the agent reads messages, reasons, may call tools, and produces output. In everyday Azure work, teams use it to track what an agent attempted, what tools it used, whether it completed, and why a response was produced. The useful evidence is thread Microsoft Foundry intermediate 4 commands Aliases: Foundry agent run, AI agent run, thread run Quick peek Open full term page

AI and Machine Learning premium Agent tool Agent tool is a capability an agent can invoke to search, run code, query data, call APIs, or perform a configured action. In everyday Azure work, teams use it to let an agent go beyond text generation by grounding answers or carrying out controlled work. The useful evidence is tool type, configuration, authentication method, allowed Microsoft Foundry intermediate 4 commands Aliases: Foundry agent tool, AI agent tool, agent built-in tool, agent custom tool Quick peek Open full term page

AI and Machine Learning premium Azure AI Search Azure AI Search is a managed retrieval service for full-text, vector, hybrid, and semantic search across application and enterprise content. It provides search services, indexes, indexers, skillsets, ranking features, security controls, and APIs used by apps, copilots, and knowledge portals. Search fundamentals 4 commands Aliases: Azure AI Search, AI Search, Azure Cognitive Search, enterprise search, vector search, hybrid search, semantic search, Search service Quick peek Open full term page

AI and Machine Learning premium Embedding An embedding is a list of numbers that captures meaning in a way software can compare. Instead of matching only exact words, an application can compare the embedding for a user question with embeddings for documents, products, tickets, or records. Similar ideas end up near each other in vector space even when the wording differs. In Azure, embeddings commonly come from Azure OpenAI models and are stored in vector indexes. Generative AI fundamentals 4 commands Aliases: Embedding, text embedding, vector embedding, embedding vector, embedding Quick peek Open full term page

AI and Machine Learning premium Embeddings Embeddings are the many vectors your system creates when it turns a collection of text, records, or documents into searchable meaning. One embedding represents one input or chunk; embeddings as a set become a retrieval layer. Applications compare a new query embedding against stored embeddings to find similar content. In Azure workloads, embeddings often connect Azure OpenAI model deployments with Azure AI Search, Cosmos DB, Azure SQL, PostgreSQL, or Redis.. Azure OpenAI fundamentals 4 commands Aliases: Embeddings, text embeddings, vector embeddings, embedding vectors, embeddings Quick peek Open full term page

AI and Machine Learning premium Fine-tuned model Fine-tuned model is a customized model produced from a supported base model by a completed fine-tuning job and then deployed for inference in Azure. Azure AI Foundry and Azure OpenAI intermediate 4 commands Aliases: Fine-tuned model, fine tuned model Quick peek Open full term page

AI and Machine Learning premium Image Analysis Image Analysis is an Azure AI Vision capability that uses pretrained models to extract visual features, text, people, objects, tags, and captions from images. Azure AI Vision intermediate 4 commands Aliases: Analyze Image, Azure AI Vision Image Analysis, Computer Vision Image Analysis, image analysis, Image Analysis Quick peek Open full term page

AI and Machine Learning premium AI evaluation AI evaluation is the Microsoft Foundry process of testing a generative AI model, agent, or application against a dataset and measuring its quality, safety, and task performance with built-in or custom evaluators. Microsoft Foundry intermediate 3 commands Aliases: Azure AI evaluation, Foundry evaluation, generative AI evaluation, model evaluation Quick peek Open full term page

Monitoring and Observability premium Application Insights connection string An Application Insights connection string specifies the Application Insights resource and endpoints that an instrumented application uses to send telemetry. Application Insights fundamentals 3 commands Aliases: APPLICATIONINSIGHTS_CONNECTION_STRING, application insights connection string Quick peek Open full term page

DevOps premium Blue-green deployment Blue-green deployment is a release strategy that runs two production-capable environments, routes users to the current version, validates the new version, then shifts traffic with minimal downtime. Deployment workflows intermediate 3 commands Aliases: No aliases yet Quick peek Open full term page

AI and Machine Learning premium BM25 BM25 is the Azure AI Search full-text relevance scoring algorithm that ranks keyword search results using term frequency, inverse document frequency, length normalization, and saturation controls. Azure AI Search intermediate 3 commands Aliases: No aliases yet Quick peek Open full term page

Databases premium Bounded staleness Bounded staleness is an Azure Cosmos DB consistency level that limits how far reads can lag writes by a configured number of versions or time interval. Azure Cosmos DB fundamentals 3 commands Aliases: No aliases yet Quick peek Open full term page

AI and Machine Learning premium Completion the generated text or structured output returned by an AI model after an application sends prompt, message, or inference request data Azure OpenAI Intermediate 3 commands Aliases: No aliases yet Quick peek Open full term page

AI and Machine Learning premium Content Understanding Content Understanding is documented by Microsoft as part of the Azure AI services area in Azure. Azure AI services intermediate 3 commands Aliases: No aliases yet Quick peek Open full term page

Databases premium Cosmos DB vector search Cosmos DB vector search is an Azure glossary term for the Azure Cosmos DB capability for storing vector embeddings with operational data and querying for similar items instead of only exact field matches. In practice, it helps teams reason about semantic retrieval, recommendation, retrieval augmented generation, and similarity search inside a Cosmos DB workload using live Azure evidence, documented ownership, and Microsoft Learn guidance. It should be reviewed through security, reliability, operations, cost, and performance lenses before production changes. Azure Cosmos DB advanced 3 commands Aliases: Azure Cosmos DB vector search Quick peek Open full term page

Monitoring and Observability premium Application map Application Map is an Application Insights view that visualizes application components and dependencies using telemetry, helping teams identify bottlenecks and failure hotspots. Application Insights fundamentals 2 commands Aliases: application map Quick peek Open full term page

AI and Machine Learning verified Prompt evaluation Prompt evaluation is how teams test whether a prompt actually works instead of relying on a good demo. You run the AI behavior against a dataset of representative inputs, expected answers, safety cases, or business rules. Evaluators score quality, grounding, safety, formatting, and task success. In Azure and Microsoft Foundry, prompt evaluation helps decide whether a prompt, model, agent, or retrieval change is ready for production. It turns subjective “looks good” reviews into measurable evidence. Azure Machine Learning intermediate 6 commands Aliases: No aliases yet Quick peek Open full term page

AI and Machine Learning verified Prompt Shields Prompt Shields is a Microsoft Foundry and Azure AI Content Safety capability that detects adversarial inputs against language-model applications. It covers direct user prompt attacks and indirect document attacks, helping apps identify attempts to override instructions, exfiltrate data, or manipulate model behavior before generation. Azure AI services intermediate 6 commands Aliases: No aliases yet Quick peek Open full term page